In the msp3 manuscript, we only included lehilahytsara samples from an area that is well north of the published IUCN range of the species, and some of which were originally labelled as mittermeieri.

So a question relevant to their (lack of) delimitation as species is what differentiation looks like when more southerly lehilahytsara populations are also included.

1. Maps

The map below shows the distributions of mittermeieri (green) and lehilahytsara (blue). IUCN distibutions have a darker color and a solid contour line; hand-drawn estimated “additional” distributions (based on our sampling sites and forest cover) have a lighter color and dotted contour lines.

The maps below shows the sites for which we have samples with RADseq data, colored by species (with remaining forest in the background). On the right, some sites that are very close to each other have been lumped together (e.g., “Anjiahely+” consists of “Anjiahely” and “Antsahabe”) – to avoid cluttering, I will use these lumped site names when visualizing PCA and ADMIXTURE results below.

Based on locations and genetic differentiation (see results below), I have divided the lehilahytsara samples into three “population groups”: “lehi-north”, “lehi-south”, and “lehi-high” (highland plateau sites, which is only Ankafobe after filtering). See the sites maps below, now colored by population group. For the msp3 paper, we only included lehilahytsara samples from the “lehi-north” population group (Riamalandy and the Ambavala area).

Finally, here is a map simply showing sampling sites by population group, with vegetation types of remaining vegetation shown:

2. Some metadata on the samples with RADseq data

Samples were genotyped with Stacks and then filtered using a fairly stringent filtering procedure which removes both sites and samples. In the tables below, the “pass” indicates samples that passed this filtering, and “fail” indicates samples that failed this filtering step. (Not all samples that “failed” this are useless though, given the stringency of the filtering.) I did not include the two or so samples from captive colonies.

##      pop_group pass fail total
## 1 mittermeieri    8    5    13
## 2  lehi: north    8    2    10
## 3   lehi: high    7    7    14
## 4  lehi: south   11    6    17
##      site_short pass fail total    pop_group
## 1      Ambatovy    4    0     4  lehi: south
## 2      Ambavala    5    1     6  lehi: north
## 3     Amboasary    1    1     2  lehi: south
## 4  Ambohitanely    0    4     4   lehi: high
## 5  Anjanaharibe    2    3     5 mittermeieri
## 6     Anjiahely    3    1     4 mittermeieri
## 7      Ankafobe    7    3    10   lehi: high
## 8       Anosibe    1    0     1  lehi: south
## 9     Antsahabe    1    0     1 mittermeieri
## 10       Madera    2    0     2  lehi: north
## 11     Mantadia    1    2     3  lehi: south
## 12     Marojejy    2    1     3 mittermeieri
## 13   Riamalandy    1    1     2  lehi: north
## 14     Sahanody    0    1     1  lehi: south
## 15  Tsinjoarivo    4    2     6  lehi: south
##         ID  Sample_ID   sp   site_short    site_lump    pop_group filtering
## 1  mmit001     MBB012 mmit Anjanaharibe Anjanaharibe mittermeieri      fail
## 2  mmit002     MBB013 mmit Anjanaharibe Anjanaharibe mittermeieri      fail
## 3  mmit003     MBB014 mmit Anjanaharibe Anjanaharibe mittermeieri      fail
## 4  mmit004     MBB016 mmit Anjanaharibe Anjanaharibe mittermeieri      pass
## 5  mmit019    PBZT115 mmit Anjanaharibe Anjanaharibe mittermeieri      pass
## 6  mmit008 04-07_hely mmit    Anjiahely   Anjiahely+ mittermeieri      fail
## 7  mmit009 07-06_habe mmit    Antsahabe   Anjiahely+ mittermeieri      pass
## 8  mmit010 08-07_hely mmit    Anjiahely   Anjiahely+ mittermeieri      pass
## 9  mmit011 10-07_hely mmit    Anjiahely   Anjiahely+ mittermeieri      pass
## 10 mmit012  2-07_hely mmit    Anjiahely   Anjiahely+ mittermeieri      pass
## 11 mmit005     MBB005 mmit     Marojejy     Marojejy mittermeieri      fail
## 12 mmit006     RMR186 mmit     Marojejy     Marojejy mittermeieri      pass
## 13 mmit007     RMR187 mmit     Marojejy     Marojejy mittermeieri      pass
## 14 mmit013        B12 mleh     Ambavala    Ambavala+  lehi: north      pass
## 15 mmit014        B14 mleh     Ambavala    Ambavala+  lehi: north      pass
## 16 mmit015        BC3 mleh       Madera    Ambavala+  lehi: north      pass
## 17 mmit016        C12 mleh     Ambavala    Ambavala+  lehi: north      pass
## 18 mmit017        C23 mleh     Ambavala    Ambavala+  lehi: north      pass
## 19 mmit018        C24 mleh       Madera    Ambavala+  lehi: north      pass
## 20 mspp016        B23 mleh     Ambavala    Ambavala+  lehi: north      fail
## 21 mspp019        BC2 mleh     Ambavala    Ambavala+  lehi: north      pass
## 22 mleh019     JMR001 mleh   Riamalandy   Riamalandy  lehi: north      fail
## 23 mleh020     JMR002 mleh   Riamalandy   Riamalandy  lehi: north      pass
## 24 mleh005      RMR95 mleh Ambohitanely Ambohitanely   lehi: high      fail
## 25 mleh006      RMR96 mleh Ambohitanely Ambohitanely   lehi: high      fail
## 26 mleh007      RMR97 mleh Ambohitanely Ambohitanely   lehi: high      fail
## 27 mleh008      RMR99 mleh Ambohitanely Ambohitanely   lehi: high      fail
## 28 mleh009     MBB036 mleh     Ankafobe     Ankafobe   lehi: high      fail
## 29 mleh010     MBB037 mleh     Ankafobe     Ankafobe   lehi: high      pass
## 30 mleh011     MBB038 mleh     Ankafobe     Ankafobe   lehi: high      pass
## 31 mleh012     MBB039 mleh     Ankafobe     Ankafobe   lehi: high      fail
## 32 mleh013     MBB040 mleh     Ankafobe     Ankafobe   lehi: high      pass
## 33 mleh014     MBB041 mleh     Ankafobe     Ankafobe   lehi: high      pass
## 34 mleh015     MBB042 mleh     Ankafobe     Ankafobe   lehi: high      pass
## 35 mleh016     MBB043 mleh     Ankafobe     Ankafobe   lehi: high      pass
## 36 mleh017     MBB044 mleh     Ankafobe     Ankafobe   lehi: high      fail
## 37 mleh018     MBB045 mleh     Ankafobe     Ankafobe   lehi: high      pass
## 38 mleh001     JMR092 mleh     Ambatovy    Ambatovy+  lehi: south      pass
## 39 mleh002     MBB001 mleh     Ambatovy    Ambatovy+  lehi: south      pass
## 40 mleh003     MBB002 mleh     Ambatovy    Ambatovy+  lehi: south      pass
## 41 mleh004     MBB003 mleh     Ambatovy    Ambatovy+  lehi: south      pass
## 42 mleh027  01-00_Man mleh     Mantadia    Ambatovy+  lehi: south      fail
## 43 mleh028  02-00_Man mleh     Mantadia    Ambatovy+  lehi: south      fail
## 44 mleh031     ODY6.9 mleh     Sahanody    Ambatovy+  lehi: south      fail
## 45 mleh033    TAD4.32 mleh     Mantadia    Ambatovy+  lehi: south      pass
## 46 mleh029     ANJZ11 mleh    Amboasary    Amboasary  lehi: south      pass
## 47 mleh030     ANJZ20 mleh    Amboasary    Amboasary  lehi: south      fail
## 48 mleh032     SIB7.1 mleh      Anosibe      Anosibe  lehi: south      pass
## 49 mleh021    DWW3235 mleh  Tsinjoarivo  Tsinjoarivo  lehi: south      pass
## 50 mleh022    DWW3236 mleh  Tsinjoarivo  Tsinjoarivo  lehi: south      fail
## 51 mleh023    DWW3243 mleh  Tsinjoarivo  Tsinjoarivo  lehi: south      pass
## 52 mleh024    DWW3244 mleh  Tsinjoarivo  Tsinjoarivo  lehi: south      pass
## 53 mleh025    DWW3249 mleh  Tsinjoarivo  Tsinjoarivo  lehi: south      pass
## 54 mleh026    DWW3250 mleh  Tsinjoarivo  Tsinjoarivo  lehi: south      fail

3. Jordi’s tree

Here is part of Jordi’s tree that he sent around a while ago. This seems to suggest that the main axis of differentiation is between mittermeieri + “lehi-north” on one hand versus “lehi-south” on the other (note that no samples from “lehi-high” were included). At its most extreme, this could mean that one could in fact distinguish two species, but that “lehi-north” would just have to be included with mittermeieri, which would make our msp3 analyses for these two species … a little off the mark.

4. Splitstree

Below is an annotated Splitstree network, with sites and population groups labelled. This is roughly in line with Jordi’s tree, but it’s fairly messy, and might rather suggest an isolation-by-distance (IBD) pattern that would show even less support for distinct groups if geographically intermediate populations were included.

5. ADMIXTURE

Below is an ADMIXTURE plot showing cross-validation errors at different values of K. This suggests that the fit gets worse with a higher number of clusters, already slightly so going from K=1 -> K=2. So no support for two species.

Below is an ADMIXTURE plot with K=2, K=3, and K=4. At K=2, “lehi-north”" is grouped with mittermeieri, like in Jordi’s tree. At K=3, “lehi-high” gets its own cluster. At K=4, each of the population groups gets its own cluster.

5. PCA

Below is a PCA plot, with ellipses around each (lumped) site, and a different point shape for each population group. The inset just shows colors by population group. I’d say these results are also consistent with a rough IBD pattern going from “lehi-south” -> “lehi-north”" -> mittermeieri, but with the highland population group relatively distinct.

7. Concluding, w/r/t/ the msp3 manuscript